NeuroSynth Compose: A Web-Based Platform for Reproducible Meta-Analyses
Poster No:
2235
Submission Type:
Abstract Submission
Authors:
James Kent1, Angela Laird2, Nicholas Lee3, Julio Peraza4, Taylor Salo5, Katherine Bottenhorn6, Jérôme Dockès7, Ross Blair8, Kendra Oudyk9, Thomas Nichols10, Jean-Baptiste Poline9, Alejandro De La Vega11
Institutions:
1Department of Psychology, University of Texas at Austin, Austin, TX, 2Department of Physics, Florida International University, , Miami, FL, 3Mcgill University, Montreal, Quebec, 4Florida International University, Miami, FL, 5University of Pennsylvania, Philadelphia, PA, 6University of Southern California, Los Angeles, CA, 7Inria, Palaiseau, Other, 8Stanford University, San Francisco, CA, 9McGill University, Montreal, Quebec, 10University of Oxford, Oxford, United Kingdom, 11University of Texas at AUstin, Austin, TX
First Author:
Co-Author(s):
Introduction:
The advent of fMRI has resulted in a deluge of over 20,000 studies mapping function to neuroanatomy. Meta-analysis is essential for gleaning insights into human brain function; however, the painstaking process of collecting, extracting, and synthesizing data is a major bottleneck limiting its routine application in scientific practice. We present a powerful and easy-to-use platform for meta-analysis, leveraging text mining, artificial intelligence and streamlined curation workflows to enable researchers to perform precise meta-analyses in a fraction of the time without leaving their browser.
Methods:
Neurosynth Compose is a modular ecosystem of tools to curate, ingest, and annotate data, and execute meta-analyses. We index over 20,000 neuroimaging studies, featuring pre-extracted imaging data from a diverse range of journals. We automatically annotate studies with ontological labels from Cognitive Atlas and Disease Ontology and extract structured meta-data such as participant count using OpenAI's GPT-3.5. We implement a web-based PRISMA curation workflow, enabling users to systematically include studies into a final set for meta-analysis, and annotate studies with custom meta-data. Finally, users can specify a reproducible meta-analysis specification, which is executed in the cloud and automatically uploads results for easy sharing.
Results:
We replicated Witt (2008) meta-analysis of 38 studies investigating finger tapping using the ALE algorithm, resulting in a quantitatively similar result to the original study. The specification and results are accessible in the unique meta-analysis page (https://bit.ly/ns-meta-analysis).
Conclusions:
Neurosynth Compose removes major barriers to meta-analysis democratizing quantitative syntheses of the vast and diverse neuroimaging literature for a wide range of researchers.
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 2
Connectivity (eg. functional, effective, structural)
Neuroinformatics and Data Sharing:
Databasing and Data Sharing 1
Workflows
Keywords:
Data analysis
Data Organization
Informatics
Open Data
Open-Source Code
Open-Source Software
Systems
Workflows
Other - meta-analysis
1|2Indicates the priority used for review
·Components of the meta-analysis workflow from curation to execution of the meta-analysis